An accurate and practical approach is necessary for predicting oil fraction in horizontal oil–water flows. In this study, a concept of a drift-flux model is adopted to develop a predictive method for the oil fraction in the horizontal oil–water flows due to its simplicity and practicality. A new drift-flux correlation for the horizontal oil–water flows is developed based on the least square method using collected experimental data. The distribution parameter is determined to be 1.05 for the data with the ratio of oil density to water density ranging from 0.787 to 1.00, whereas the oil fraction weighted mean drift velocity is set at 0 m/s due to the flow direction perpendicular to the gravity direction. The physical meaning for the order of unity of the distribution parameter is explained by introducing a simple model. The predictive capability of the new drift-flux correlation is examined using the collected database of oil–water flows in horizontal pipes under a variety of test conditions. It is demonstrated that the new drift-flux correlation can predict the existing oil fractions in the horizontal pipe channels with the mean absolute error, standard deviation, mean relative deviation, and mean absolute relative deviation being −0.0124, 0.0338, −3.25%, and 9.57%, respectively.